The disclosure relates generally to training set creation for classifying features of a system under agile development.
In general, contemporary implementations of software development include a feature analysis tool to select and identify top risk items. The feature analysis tool can use finite state of a project and features, which identify whether the project is done and whether a state of each feature is known (e.g., correct or incorrect). If a defect in a particular feature is found, that feature is assigned to a ‘BAD’ class. Otherwise, that feature is assigned to an ‘OK’ class. However, because the features analysis tool only monitors a completed project, the feature analysis tool cannot identify when internal feature iterations fail. This is not acceptable for software development under an agile development model, which requires constant testing in parallel to development.